SETTLE_DEFAULTS

const SETTLE_DEFAULTS: SettleOptions

Defined in: packages/runtime/src/settle.ts:63

Defaults measured against the corpus rather than chosen: 0.5 s resolves a 1.5 s settle without splitting hairs, and 0.25 dB is below the run-to-run spread.

The top rung is deliberately above anything the corpus reaches, and that is the point. It was 16 s, justified as “covers marshall-jcm800, the slowest row found (settles at 16.0 s)” — which is not a bound, it is a coincidence. A ladder whose maximum equals the population maximum cannot tell settled at the last rung from never settled: both produce the same row, and jcm800’s gain of 0.00 — the packet called the one real remaining amp defect all afternoon — was read at exactly that edge every time.

Re-measured on a 64 s ladder, jcm800 settles at 10 s and holds flat to five digits out to 64 s, so the diagnosis survives and the 16.0 s figure was an artifact of the two settling criteria this module used to have (see windowLevel). The rung is raised anyway: the reason to have headroom is not that anything needed it, it is that without it a maximum reading is ambiguous. Same argument as a screen needing a no valid measurement state distinct from a negative result. It costs nothing for a packet that settles earlier.